Advances in Computer Science Research, volume 87
3rd International Conference on Mechatronics Engineering and Information Technology (ICMEIT 2019)
Simultaneous Localization and Mapping of Mobile Robot with
Research and Implementation
Chen Du1, a and Yu Du2, b, *
1Beijing Key Laboratory of Information Service Engineering, Beijing Union University, Beijing
100101, China;
2 College of Robotics, Beijing Union University, Beijing 100101, China.
axueshdch@163.com, b, * duyu@buu.edu.cn
Abstract. In the process of autonomous navigation, mobile robots need to build maps of the
surrounding environment and simultaneous localization. The Rao-Blackwellzed particle filter
algorithm is one of the methods to efficiently solve the problem that simultaneous localization and
mapping of mobile robots. At present, Mapping of inconsistent have long been the focus of research.
In order to solve this problem, this paper provides an algorithm which uses high-precision Laser data
to correct the proposed distribution based on odometer readings, focus sampling on the possible
areas of observation information, reduces the error of proposed distribution, and establish a more
accurate map environment. Finally, the experimental verification was carried out on the Bulldog
mobile robot platform equipped with a 16-line Laser sensor. The results show that the optimized
method of performance is more stable, can improves the diversity of particles and creates high-
precision environmental maps online in real time.
Keywords: Sensor technique; Simultaneous localization and mapping; Rao-Blackwellized particle
filter; Robot operating system.
1. Introduction
With the rapid development of science and technology society and the vigorous development of
artificial intelligence, mobile robot technology, which is condensing many scientific and
technological achievements such as control engineering technology, computer technology, sensor
technology, material science, mechanical manufacturing technology, signal processing, has entered a
new upsurge. It is an important basis and symbol for intelligence to be able to identify and locate the
environment autonomously, so as to be able to move, judge and act autonomously. Therefore, among
the many research of mobile robots, simultaneous localization and mapping (SLAM) has become a
very important part of many research [1, 3].
Simultaneous Localization and Mapping consists in the simultaneous estimation of the state of the
robot and the map of the environment according to the environmental characteristics observed by its
sensors, such as Laser and odometer. It includes simultaneous estimation of robot state and
environment map [4]. In simple instances, the robot state is the representation of its position, direction
and speed, which can also include other quantities, such as sensor biases and calibration parameters.
The map, on the other hand, is a representation of robot operating environment, such as landmark
position, obstacles and so on. The establishment of high-precision maps facilitates the execution of
other robot tasks and is a very important part of the research of mobile robots.
After many years, many improved algorithms about RBPF-SLAM have emerged. Wang Tian
Orange [5] introduced a kind of region Particle Swarm Optimization (PSO) method into RBPF-
SLAM algorithm to adjust the particles’ proposal distribution. All particles are clustered into several
regions and the weighted central position of each region is calculated. Luo Yuan [6] designed an
RBPF-SLAM algorithm based on annealing parameters to optimize the hybrid proposed distribution.
The annealing parameters were used to control the proportions of the mixed proposed distributions,
so that the improved proposed distribution estimates were closer to the real state. Goran [7] proposed
a grid-based modification method of Rao-Blackwellized particle filter, which relies on the floor plan
of the indoor environment to provide a priori information for simultaneous localization and mapping,
so as to have stronger robustness.
Copyright © 2019, the Authors. Published by Atlantis Press.
This is an open access article under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/). 577
Advances in Computer Science Research, volume 87
Based on the Rao-Blackwellized particle filter method, this paper uses high-precision Laser data
to correct the proposed distribution based on the odometer readings. In order to verify the
effectiveness of the algorithm, based on the Bulldog mobile robot platform, relevant experiments
were carried out.
2. SLAM based on Rao-Blackwellized
The problem of Simultaneous localization and mapping can be understood as: when the initial
state of the robot is known, that is, the initial map m0 and the initial pose x0 , the sensor observation
information z1: t z1 ,, zt from the initial to the moment, and the control information u1: t 1 u1 ,, ut 1
of the odometer, the joint posterior probability p( x1: t , m | z1: t , u1: t 1 ) of the robot's trajectory
x1: t x1 ,, xt and the environmental map mt can be estimated. As shown in formula 1:
p( x1: t , m | z1: t , u1: t 1 ) p(m | x1: t , z1: t ) p( x1: t | z1: t , u1: t 1 ) (1)
This formula divides the joint posterior distribution into products of two posterior distributions.
On the one hand, the posterior probability estimation of the robot's posture is obtained by collecting
and processing the observation and control information of each particle; on the other hand, the
posterior probability estimation of the map is obtained under the condition that the robot's posture
estimation and the observation information of external sensors are known.
The process can be summarized by the following four steps [8]:
1) Sampling: According to the proposed distribution, and obtaining the next generation particle set
Xt(i) from the previous generation particle set Xt(i1) . The motion model of the system is typically used
as a proposal distribution.
2) Weight: In order to better estimate the proposal distribution and reduce the error with the target
distribution, the importance weight of each particle needs to be calculated, as shown in formula 2:
p( x1(:i t) | z1: t , u1: t 1 )
t(i) (2)
q ( x1(:i t) | z1: t , u1: t 1 )
3) Resampling: Resampling the particles of the system according to the set threshold. After
resampling, all the particles have the same weight.
4) Map Estimation: For each particle, its motion trajectory x1(:it) and observation information z1: t
can be used to calculate the corresponding map p(m (i ) | x1(:it) , z1: t ) .
3. Implementation based on Optimized RBPF-SLAM
ROS (Robot Operating System), a robot operating system, is a distributed processing framework.
ROS are divided into: node, manager, message, and topic. The core is a node that is used to perform
a relatively simple task or process. Multiple nodes pass messages to each other and can independently
initiate or terminate. A node can issue messages or provide services for specific titles to other nodes
[9].
All experiments were performed on the ROS system, and the main node simplified diagram is
shown in Fig.1.
Fig. 1 Main nodes running on the ROS
578
Advances in Computer Science Research, volume 87
The odometer node, the laser node and the control node publish messages to the base_link topic,
the base_link interacts with the message, the information issued by the control output node is used to
control the movement of the robot, and the mapping node extracts the message of the base_link topic
for map construction.
The experimental platform is a Bulldog robot with a length of 90cm, a width of 70cm and a height
of 80cm. The picture is shown in Fig. 2. The robot is equipped with a Rslidar laser sensor and an
internal PC connected by a serial cable. The internal PC is a ROS system running on Linux (Ubuntu
14.04), which enables real-time online map construction. The external PC connects to the Bulldog
route and communicates via the HTTP protocol. In addition, this experiment also uses a handle for
controlling the Bulldog robot during the map construction process.
Fig. 2 Picture of mobile robot
First, the experimental scene is approximately 2.5m × 5m. Figure 3 shows the RBPF-SLAM
algorithm using only the odometer motion model and the high-precision laser data to correct the high-
precision map based on the proposed distribution of the odometer readings.
(a) (b)
Fig. 3 Mapping with different proposed distributions
As time goes on, the error of odometer becomes larger and larger, and the phenomenon of map
inconsistency becomes more and more obvious, as shown in the edge of in Figure 3 (a). Point A, B,
and C in Figure 3 (b) can accurately reflect the map environment.
Second, the approximate map is 6m x 8m, and there are some obstacles, such as desks and chairs,
bookcases, long tables, etc. The traditional RBPF-SLAM algorithm and the optimized RBPF-SLAM
algorithm can clearly see the outline of points A, B and C in Figure 4 by using different particle
numbers.
B
A
C
Fig. 4 Experimental scenario
579
Advances in Computer Science Research, volume 87
Table 1 shows the two algorithms using different particle numbers to construct the same sharpness
map and compare the number of particles.
Table 1. Comparison of particle numbers
Numble Algorithm Particle Numbers
1 Traditional Algorithm 38
2 Optimization Algorithm 18
As can be seen from the results of the construction of the map, when the experimental environment
is more complicated, it can be obtained from Table 1. When constructing the map with the same
definition, the optimized RBPF algorithm uses fewer particles.
4. Summary
Optimizing the Rao-Blackwellized particle filter of simultaneous localization and mapping, using
high-precision laser data to correct the proposed distribution based on odometer readings, reducing
the error of the proposed distribution, and improving the accuracy and effectively reducing the map
inconsistency. Further, the optimized RBPF algorithm uses fewer particles, and it can be seen that the
performance of the optimization algorithm is better.
References
[1]. R. Smith, M. Self, P. Cheeseman. Estimating Uncertain Spatial Relationships in Robotics[J].
Machine Intelligence & Pattern Recognition, 2013, 5(5):435-461.
[2]. G. Bresson, Z. Alsayed, L. Yu, et al. Simultaneous Localization and Mapping: A Survey of
Current Trends in Autonomous Driving[J]. IEEE Transactions on Intelligent Vehicles, 2017,
2(3):194-220.
[3]. A. A. Panchpor, S. Shue, J. M. Conrad. A survey of methods for mobile robot localization and
mapping in dynamic indoor environments[C]// 2018 Conference on Signal Processing and
Communication Engineering Systems (SPACES). IEEE, 2018.
[4]. C. Cadena, L. Carlone, H. Carrillo, et al. Simultaneous Localization and Mapping: Present,
Future, and the Robust-Perception Age[J]. IEEE Transactions on Robotics, 2016, 32(6).
[5]. T. CH. Wang, Y. F. CAI, Z. M. Tang. SLAM Method Based on Region Particle Swarm Optimi
zation and Partial Gaussi an Resampling[J]. Computer Engineering, 2017, 43(11):310-316(in
Chinese).
[6]. Y. Luo, Q. Sun, Y. ZHang, et al. simultaneous localization and mapping implementation based
on optimized RBPF[J]. Journal of Huazhong University of Sci & Tech (Natural Science), 2016,
44(5): 30-34(in Chinese).
[7]. G. POPOVIć, J. ORšULIć, D. MIKLIć, et al. Rao-Blackwellized Particle Filter SLAM with Prior
Map: An Experimental Evaluation[C]// Iberian Robotics conference. Springer, Cham, 2017:14-
25.
[8]. G. Grisetti, C. Stachniss, W. Burgard. Improved Techniques for Grid Mapping with Rao-
Blackwellized Particle Filters[J]. IEEE Transactions on Robotics, 2007, 23(1):34-46.
[9]. A. Koub. Robot Operating System (ROS) - The Complete Reference (Volume 1) [M]// Robot
Operating System (ROS): The Complete Reference (Volume 1). Springer Publishing Company,
Incorporated, 2016.
580
